In the previous work, we developed a dorsiflexion support robotic technology (RT) to help with gait rehabilitation for hemiplegic patients. The RT actively supports dorsiflexion force during swing phase with an artificial muscle. Because of the insufficient support of heel rocker, we consider using a spring for assisting resistive dorsiflexion force during loading response phase. The spring can be changed easily on our developed RT, which contributes to advantage of choosing spring with suitable spring constant. Thus, the ultimate goal for us is to find a way to determine the appropriate spring constant for corresponding individual gait to support full heel rocker in loading response phase. Support amount of dorsiflexion torque is needed for determining the spring constant. Therefore, our aim in this paper is to find out the correlation between ankle torque and individual gait factors, such as walking speed and step length. After selecting these factors, we collected gait data for 9 healthy volunteers and derive the formula to estimate dorsiflexion torque using multiple linear regression analysis. An accuracy evaluation test was also done to prove the availability of derived estimation formula.